Probably one of the difficult parts of geometric modeling is where and how to begin the process. As with most procedures, the first step should consist of a review, analysis, and planning phase. Hopefully, some of what has been presented up to this point will help in this initial phase. The modeler should think about the task that lies ahead, considering such things as the following: just exactly what is the final expected product; what will the model be used for; how much descriptive information is available, and what form is it in; what, if anything, has been done in the past; and can existing ("library") components be used. Next, one should devise a general plan on how to accomplish the actual construction of the model, including rough time estimates for each phase. This general plan should include all the major subsystems of the target that must be modeled. For most military targets, the tersest general plan would consist of the following: (1) model the exterior shell, (2) model the internal air, and (3) model all the remaining components. Of course we are being facetious about step (3), but steps (1) and (2) of this general plan are genuine. The goal is to get a "correct" exterior shell (with internal air if necessary) before any components are added. If a tank were being modeled, the first steps might become as follows: model the hull shell, model the hull air, model the turret shell, and model the turret air.
Finally, since much of the modeling effort goes into step (3) (model all remaining components), we will discuss how to model components. This discussion will be more "cookbook" oriented, but we still will not go into the details. Other publications  have discussed geometric modeling with BRL-CAD in detail. As always, initially, time should be spent analyzing and planning. First analyze the component to be modeled and decide on the detail required, formulating exactly how to represent the component, including what solids to use and how to combine them into regions. Then create the solids, so they have the desired shape, size, and orientation. To take advantage of any symmetry, create the solids at the origin. Next combine the solids into the regions, then group these regions into a combination representing the completed component. At this point check for and fix any interferences (regions occupying the same volume) within the component and create pictures for visual verification. Note that both of these actions are performed using ray-tracing. The model of the component is now completed, so it's time to move it to its location within the target. This task is accomplished by editing the combination (translate, scale, and/or rotate). Recall, this editing is stored in the matrices of the member records, so the components primitives (solid records) are still at the origin. It is recommended to "push" (an MGED command) this editing down to the solid records at this time. Next check for and fix any interferences between this component at its new location and the rest of the description. This process is repeated for every component that is to be modeled.